Community
Participate
Working Groups
Here's as much info from the series of events as I can remember: * tried to submit a quick update to bug#161646 * clicked submit to repository * some time passed, and then i got an error (see PNG attachment to come) * the issue in the task list still has the error for pending outgoing changes * the issue remained opened, but lost it's "*" at the top, meaning the system recognized it as having been saved * i closed it and reopened it, and it looked the same - no "*", my comment was still in the "new comment field", and the "submit to repository" button was not grayed out * i right-clicked on the item in the task list and selected "synchronize" * that immediately made the pending outgoing changes arrow disappear * i closed it and reopened it, and now my update was posted at the 8th comment (as opposed to still being in the new comments area) Other tidbits that might be of interest: * my internet connection was stable at the time * my IP wasn't changing / didn't change while this happened.
Created attachment 54297 [details] Report Submission Error The text body says "IO Error: Read Timed Out", but I was really trying to write my recent comment to the repository. Are there any logs I can peruse to hopefully find / provide more concrete evidence as to what precisely happened?
The bugs.eclipse.org server has been really slow resulting in time out when reading responses. After submission we re-read the report (which is why you see a read timeout). We should be able to improve things by leaving the task in a synchronized state upon successful submission (regardless of successful sychronization). Although the side effect would be an incoming notification later if the task wasn't successfully synched upon submission.
That wouldn't be too bad. I don't mind seeing a delayed notification as opposed to an error message that I'm unsure about. Error messages tend to make me want to click here or try something there to make sure my issue update really went through. This may even result in duplicate submissions (I speak from experience, see bug#165385). So, if there is *any* sort of decent and, moreover, silent error handling we can do, great. Perhaps we can even handle this error with a generic, but pleasant notification: "Bug <so-and-so> had synchronization issues. You may close the issue as it has been successfully saved. We will continue retrying the synchronization in the background. Upon successful synchronization, you will be notified as usual." Thoughts?
Yes, I think as long as the data made it to the repository, we should fail silently (erring on the side of false incoming). If this doesn't work or we are unable to guarantee successful commit we may need a message as you suggest.
Also, since bugs.eclipse.org is super slow right now, I was able to twice reproduce the IO error in this case. Each time this happens, the issue stays open and is still prefixed with "*<unsubmitted>" even when it eventually makes it to the server. So, this just happened to me again while trying to file 165392. I also inadvertently created 165396. Thus, if anything is done about this, the "*unsubmitted..." issue should be closed to help prevent the possibility of double submissions during times of high load when IO issues like this are bound to happen more frequently.
Improved new bug submission. IF the submission is successful but we are unable to open/synch (which is has led to your duplicates) we now display a dialog informing that you will need to open via Navigate > Open Repository Task... and provide the new bug id (editor is closes as you suggested). This scenario shouldn't be the norm though now that eclipse.org's bugzilla is improving. This fix will be available in tonight's release. Regarding the ugly and uninformative Read Timed Out, any further improvement will be post 1.0 (although such events should be less problematic, now that the task state is updated immediately after the post rather as part of the following synchronization). Let me know how things are performing for you if you get a chance. Will leave at P2 so we don't let this slide.
Marking fixed. - Further bug submission improvements: bug#165498 - Error handling improvements: bug#165383, bug#168049
